refactor: converge Maestro input handling (#623)

* refactor: converge Maestro input handling

* fix: preserve Maestro optional input semantics

Author: thymikee

src/compat/maestro/__tests__/replay-flow.test.ts

@@ -90,7 +90,7 @@ env:
   assert.equal(parsed.actions[3]?.flags.intervalMs, 150);
   assert.equal(parsed.actions[4]?.flags.holdMs, 3000);
   assert.equal(parsed.actions[1]?.flags.maestro?.allowNonHittableCoordinateFallback, true);
-  assert.equal(parsed.actions[6]?.flags?.maestro?.allowNonHittableCoordinateFallback, undefined);
+  assert.equal(parsed.actions[6]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
 });
 
 test('parseMaestroReplayFlow maps iOS openLink through the app id when available', () => {
@@ -223,7 +223,7 @@ test('parseMaestroReplayFlow keeps focused inputText and pressKey Enter as separ
   assert.deepEqual(parsed.actionLines, [3, 4, 5]);
 });
 
-test('parseMaestroReplayFlow marks tapOn before inputText for snapshot tap focus', () => {
+test('parseMaestroReplayFlow keeps tapOn inputText without Enter on Maestro path', () => {
   const parsed = parseMaestroReplayFlow(`appId: com.callstack.agentdevicelab
 ---
 - tapOn:
@@ -238,7 +238,29 @@ test('parseMaestroReplayFlow marks tapOn before inputText for snapshot tap focus
       ['type', ['Saved list']],
     ],
   );
-  assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, undefined);
+  assert.deepEqual(parsed.actionLines, [3, 5]);
+  assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
+});
+
+test('parseMaestroReplayFlow preserves optional tapOn before inputText without Enter', () => {
+  const parsed = parseMaestroReplayFlow(`appId: com.callstack.agentdevicelab
+---
+- tapOn:
+    id: editableNameInput
+    optional: true
+- inputText: Saved list
+`);
+
+  assert.deepEqual(
+    parsed.actions.map((entry) => [entry.command, entry.positionals]),
+    [
+      ['__maestroTapOn', ['id="editableNameInput"']],
+      ['type', ['Saved list']],
+    ],
+  );
+  assert.deepEqual(parsed.actionLines, [3, 6]);
+  assert.equal(parsed.actions[0]?.flags?.maestro?.optional, true);
+  assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
 });
 
 test('parseMaestroReplayFlow coalesces tapOn inputText while preserving pressKey Enter submit', () => {
@@ -281,6 +303,27 @@ test('parseMaestroReplayFlow does not coalesce text entry for non-input-looking
   assert.equal(parsed.actions[0]?.flags?.maestro?.allowNonHittableCoordinateFallback, undefined);
 });
 
+test('parseMaestroReplayFlow maps focused input commands to native type and keyboard actions', () => {
+  const parsed = parseMaestroReplayFlow(`appId: com.callstack.agentdevicelab
+---
+- inputText: hello
+- eraseText:
+    charactersToErase: 3
+- pasteText: pasted
+- pressKey: Return
+`);
+
+  assert.deepEqual(
+    parsed.actions.map((entry) => [entry.command, entry.positionals]),
+    [
+      ['type', ['hello']],
+      ['type', ['\b'.repeat(3)]],
+      ['type', ['pasted']],
+      ['__maestroPressEnter', []],
+    ],
+  );
+});
+
 test('parseMaestroReplayFlow rejects relative runScript paths without source path', () => {
   assert.throws(
     () =>

src/compat/maestro/replay-flow.ts

@@ -104,19 +104,12 @@ function optimizeTypedAfterTap(
   actionLines: number[],
   index: number,
 ): { actions: SessionAction[]; actionLines: number[]; consumed: number } | null {
-  const action = actions[index]!;
-  const nextAction = actions[index + 1];
-  const typedAfterTap = readPlainTypeText(nextAction);
-  const tapSelector = readPlainMaestroTapSelector(action);
-  if (!nextAction || typedAfterTap === null || tapSelector === null) return null;
-  const line = actionLines[index] ?? 1;
+  const candidate = readTypedAfterTapCandidate(actions, actionLines, index);
+  if (!candidate) return null;
+  const { action, nextAction, pressEnterAction, tapSelector, typedAfterTap, line } = candidate;
   if (!isLikelyTextEntrySelector(tapSelector)) {
     return { actions: [clearMaestroNonHittableTap(action)], actionLines: [line], consumed: 1 };
   }
-  const pressEnterAction = actions[index + 2];
-  if (pressEnterAction?.command !== MAESTRO_RUNTIME_COMMAND.pressEnter) {
-    return { actions: [clearMaestroNonHittableTap(action)], actionLines: [line], consumed: 1 };
-  }
   return {
     actions: [
       {
@@ -137,6 +130,36 @@ function optimizeTypedAfterTap(
   };
 }
 
+function readTypedAfterTapCandidate(
+  actions: SessionAction[],
+  actionLines: number[],
+  index: number,
+): {
+  action: SessionAction;
+  nextAction: SessionAction;
+  pressEnterAction: SessionAction;
+  tapSelector: string;
+  typedAfterTap: string;
+  line: number;
+} | null {
+  const action = actions[index]!;
+  const nextAction = actions[index + 1];
+  const pressEnterAction = actions[index + 2];
+  if (pressEnterAction?.command !== MAESTRO_RUNTIME_COMMAND.pressEnter) return null;
+  if (action.flags?.maestro?.optional === true) return null;
+  const typedAfterTap = readPlainTypeText(nextAction);
+  const tapSelector = readPlainMaestroTapSelector(action);
+  if (!nextAction || typedAfterTap === null || tapSelector === null) return null;
+  return {
+    action,
+    nextAction,
+    pressEnterAction,
+    tapSelector,
+    typedAfterTap,
+    line: actionLines[index] ?? 1,
+  };
+}
+
 function clearMaestroNonHittableTap(action: SessionAction): SessionAction {
   const maestro = { ...(action.flags?.maestro ?? {}) };
   delete maestro.allowNonHittableCoordinateFallback;

src/daemon/handlers/__tests__/session-replay-vars.test.ts

@@ -27,7 +27,7 @@ const LOC = { file: 'test.ad', line: 1 };
 type CapturedInvocation = {
   command: string;
   positionals?: string[];
-  flags?: Record<string, unknown>;
+  flags?: CommandFlags;
 };
 
 async function runReplayFixture(params: {
@@ -1247,7 +1247,7 @@ test('runReplayScriptFile lets snapshot id tap handle Maestro one-point edge con
   );
 });
 
-test('runReplayScriptFile keeps Maestro text-entry tapOn on the snapshot path', async () => {
+test('runReplayScriptFile coalesces Maestro text-entry tapOn into native fill', async () => {
   const calls: CapturedInvocation[] = [];
   const { response } = await runReplayFixture({
     label: 'maestro-tap-input-text-snapshot',
@@ -1257,6 +1257,7 @@ test('runReplayScriptFile keeps Maestro text-entry tapOn on the snapshot path',
       '- tapOn:',
       '    id: editableNameInput',
       '- inputText: Saved list',
+      '- pressKey: Enter',
       '',
     ].join('\n'),
     flags: { replayBackend: 'maestro' },
@@ -1284,12 +1285,12 @@ test('runReplayScriptFile keeps Maestro text-entry tapOn on the snapshot path',
   assert.deepEqual(
     calls.map((call) => [call.command, call.positionals]),
     [
-      ['snapshot', []],
-      ['click', ['120', '120']],
-      ['type', ['Saved list']],
+      ['wait', ['id="editableNameInput"', '30000']],
+      ['fill', ['id="editableNameInput"', 'Saved list']],
+      ['keyboard', ['enter']],
     ],
   );
-  assert.equal(calls[0]?.flags?.noRecord, true);
+  assert.equal(calls[1]?.flags?.maestro?.allowNonHittableCoordinateFallback, true);
 });
 
 test('runReplayScriptFile resolves Maestro swipe.label from a labeled element rect', async () => {